Dairy cows real time behavior monitoring by energy-efficient embedded sensor

Monitoring the behaviors of dairy cows has the potential to improve their health, welfare and productivity. Therefore, sensors attached to their body parts (neck, leg and back, etc) are useful to quantify these behaviors. Indeed, numerous sensors are used to predict diseases, stress, etc. However, they are restricted by constraints such as their size and the power consumption. In this study, we propose a new non-invasive and energy-efficient sensor to monitor and classify in real time the dairy cow's behaviors. It uses an accelerometer to track the inclination of the dairy cows' backs. This sensor detects cow motion activities and transition periods between standing and lying. To reduce power consumption, a new data selection method is integrated in the sensor to reduce the data before performing the classification. Moreover, a new time-driven technique based on sleep/wake-up methods is adopted. The results show an accuracy of 100% in transition detection with a data reduction of 99.2% and the approximate power consumption of the sensor is 0.043 mA.